Weekly Assignment #7

Michael Patel

import numpy as np
import pandas as pd
import altair as alt
policeData = pd.read_csv("Demographics_of_Police_Stops.csv")
policeData
# this data includes the date, time, and location of stops, along with sex, race, ethnic, and yearOfBirth.

	ObjectId	stopdate	stoptime	streetnbr	streetdir	street	Min_	sex	race	ethnic	Year_of_birth	enfaction	rpmainid
0	1	2018/01/19 00:00:00	30	NaN	NaN	ARAPAHOE AVE/CANYON BLVD	15	M	W	N	1959	N	3174
1	2	2018/01/13 00:00:00	2235	2398	NaN	CANYON BLVD/FOLSOM ST	10	M	W	N	1989	Y	3175
2	3	2018/01/19 00:00:00	149	1849	NaN	11TH ST/CANYON | WALNUT ALLEY	5	M	W	N	1996	Y	3176
3	4	2018/01/19 00:00:00	149	1849	NaN	11TH ST/CANYON | WALNUT ALLEY	5	M	W	N	1996	Y	3177
4	5	2018/01/11 00:00:00	2300	2299	NaN	30TH ST/PEARL ST	5	F	W	N	1998	Y	3178
...	...	...	...	...	...	...	...	...	...	...	...	...	...
8210	8211	2018/12/14 00:00:00	2245	398	NaN	S BROADWAY/DARTMOUTH AVE	10	M	W	N	1981	Y	10930
8211	8212	2018/12/15 00:00:00	1848	1599	NaN	FOLSOM ST/ARAPAHOE AVE	10	F	W	N	1976	Y	10931
8212	8213	2018/12/15 00:00:00	1749	2900	NaN	WALNUT ST	3	M	W	N	1983	N	10932
8213	8214	2018/12/15 00:00:00	1320	5699	NaN	VALMONT RD/N 57TH ST	10	M	W	N	1971	Y	10933
8214	8215	2018/12/15 00:00:00	1405	599	NaN	27TH WAY/MOORHEAD AVE	10	F	W	H	1994	Y	10934

8215 rows × 13 columns

As we can see from the dataset's documentation:

This data contains information related to officer-initated stops by the City of Boulder Police Department. Information on the demographics of the person stopped (sex, race, ethnicity, year of birth, whether they are a Boulder resident) is included in this file. See the "Outcomes of Police Stops" dataset for more details on the outcome of the stop (stop location, duration, search, and result). This demographic data is collected at the stop level, and no individual-level identifiers are recorded in the system during a stop.

The data published are limited to stops where the officer initiated, or had discretion, in making a stop. Instances where an officer is responding to a community or police call are considered non-discretionary, and demographics information is not collected for those stops and not included here. There are some instances of non-discretion within a stop interaction as well. For example, there may be instances where there is an outstanding felony warrant for the person stopped, and by law the officer must arrest that person.

Version 1

alt.data_transformers.disable_max_rows()
raceHist = alt.Chart(policeData).mark_bar().encode(
            x="race",
            y="count()")
raceHist

print("White stops: " + str((7432/8215)*100)+"%")
print("Black stops: "+str((353/8215)*100) +"%")
print("Asian stops: "+str((310/8215)*100)+"%")

White stops: 90.46865489957395%
Black stops: 4.2970176506390745%
Asian stops: 3.7735849056603774%

As we can see above, the histograms of both Race and Sex are plotted. We see quite a drastic difference in both, where White and Males were both stopped more frequently. While the evidence seems quite clear that White people are being stopped drastically more often than any other race, it is important to consider the City of Boulder's overall demographic. Boulder is almost 90% White, while Asians represent almost 6% of the population, and African Americans only a little more than 1%. When taking this into consideration, we see that White people are stopped in 90.5% of the incidents, while black people are stopped in 4.3% of all stops in 2018. Without any further statistical analysis, we cannot draw any conclusions from this, however, it is pretty clear that White's representation in police stops is pretty accurate to their actual representation of the population. At the same time, black people are stopped in about 4.3% of all stops, while only representing about 1.2% of the population. Again, this is not a statement of cause and effect, but rather that the data suggests racial discrimination against African Americans.

Version 2

Now that we have a distribution of stops by race, let's make it more readable and interesting.

alt.Chart(policeData).mark_bar().encode(
    x="count()",
    y="race",
    color=alt.Color("sex", scale=alt.Scale(scheme="dark2"))).properties(
    width=800,
    height=300)

Version 2 of my graph is now larger and horizontal, making the relative differences much clearer. Previously, it was difficult to see the difference between Asian and Black stops, for example, because they are relatively rare compared to the White bar. Additionally, the bar is now stacked, showing the breakdown of Sex, which is represented with a clear color scheme that is easily decipherable.

Version 3

alt.Chart(policeData).mark_bar().encode(
    x="count()",
    y="race",
    color=alt.Color("sex", scale=alt.Scale(scheme="dark2")),
    tooltip=["count()","sex"]).properties(
    width=800,
    height=300)

Please hover your mouse over each bar and each section. Here, you should see the Count of Records, which indicates how many stops were made for that specific Race and Sex. It seems that there is quite a difference in the stoppages regarding Sex; Males are stopped more often for each Race.